Cultural models in

language and thought

EDITED BY

Dorothy Holland
Universily oj North Carolina, Chape/ Hill

Naomi Quinn
Duke University

(lw_,l,,
Tlu "lltt•f,Jt,
.¡c.,.,;,)l¡r
1• 1'1111WUII
~ ...,._..,.,....,.,

_ J , ,.,.urJt ¡
,,H-1 • m .. IJu
u.,.,,.,,~,,_,,.¡

flld~l i,.,J ,_.........,,

,,., IJU ,

Cambridge University Press

CAMBRIDGE
NEW YORK NEW ROCHELLE
MELBOURNE SYDNEY
2
The dejinition oj lie
AN EXAMINATIO N OF THE FOLK MODELS
1
UNDERLYING A SEMANTIC PROTOTYP E

Eve E. Sweetser

This paper investigates how the semantic structure of one English word
depends on, and reflects, our models of relevant arcas of experience. As
a linguist, my original concern was with the problems posed by the word
/ie for traditional semantic theories; but these problems led inexorably to
the cultural models of informational exchange that moti vate the existence
of a sell)antic entity mcaning lie. 2 1 begin by posing the semantic problem
and go on to the cultural solution.
George Lakoff (1972), Fillmore (1977), and Coleman and Kay (1981 )
bave argued against traditional generativc and structuralist "ehccklists"
of semantic features that constitute necessary and sufficient conditions
for set-membership in the category denoted by a word. Lexical categories
3
can have better or worse members, or partía/ members. Kay and
McDaniel (1978) have shown that color categories lack necessary and suf-
ficient conditions; red is a gradicnt quality whosc category-boundaries
are best described by fuzzy set theory rather than by tradiúonal sct theory.
Checklist feature-definitions, which do not allow for color's being "sorta
red," must be replaced by a thcory capable of dealing with fuzzy set-
mcmbership. Prototypc semantics views word- mcaning as determined by
a centraJ or prototypical application, rather than a category-boundaries.
Clear definitions can thus be given for words with fuzzy boundaries of
application. We define the best instance of a word's use, and expect real-
world cases to fit this best example more or less, rather than perfectly or
not at all.
Coleman and Kay ( 1981) show that prototype theory is needed to ex-
plain the usage of thc verb /ie} As is natural in·prototype semantics (but
not in traditional set-membership semantics), Jying is a matter of more
or less. Clear central cases of Jies occur when all of Coleman and Kay's
proposed conditions are fulfilled; namely, (a) speaker believes statement
is false; (b) speaker intends to deceive hearer by maldng the statement;
and (e) the statement is false in fact. Conversely, a statement fulfilling
none of a-e is a clear nonlie. But when only one or two of a-e hold,
speakers are frequently confused and find it difficult to categorize an ac-
tion as líe or nonlie. Furthcr, these conditions (unlikc checklist-featurcs)
 EVE E. SWEETSER THE OEFINITION OF LIE 45
44

differ in weight, (a) bcing strongcst and (e) weakcst in influencing spcakers' happy, or negotiating the interaction-frame, may be a more important
categorization of acts as líes. goal than informativeness. The maxim of informationality is thus binding
Prototype semantics has been attenúve to the grounding of language precisely to the degree that we consider oursclves to be operating in a
in the speaker's world. Kay and McDaniel found physi~ pcrccptual simplified world in which discourse is informational, so that the default
reasons for color-tcrm universals; Rosch (1978) and Merv1s and Rosch purpose of an utterance is not joking, politeness, or frame-bargaining.
(1981) demonstrate that linguistic categories depend on general_ hu~an Our covert discourse-purposes are only made possible by a cultural model
category-formation abilities. Fillmore ( 1977) discusses sorne ways m whtch that cstablishes our overt purpose as informational; frame-bargaining,
the social world shapes word-mcaning. Bachelor is a classic di fficult case: and most indirect speech, depend on having "direct" speech say something
Why is it difficult to say whether thc Pope, or a thrice-married divorcé, else.
can be called a bachelor? The answer, Fillmore says, is that bachelor I sketch sorne rclevant aspects of our folk understanding of informa-
depends on a simplified world-view in which people are marriageable at tionallanguage-use and then use this cultural model to explain the rnean-
a certain age, mostly marry at that age, and stay rnarried to t~1e sarne ing of /ie as presentcd by Coleman and Kay. First, Jet us posit two basic
spouse. In this simplified world, a bachelor is simply any unmarned male principies as parts of our model of general social interaction rather than
past marriageable age; outside the sim~lified world, the wo~d bacheior of our specific model of speech acts. These principies, which are assumed
just does not apply. Bachelor necessanly evo~es a. prototyptcal schema to opera te in thc dcfault case (like Griccan maxims), are (1) Try to help,
of marriage within our cultural model of a hfe-htstor~. . . not harm and (2) Knowledge is beneficia.l. Togethcr, the two principies
¡ argue that like bachelor, lie is inherently grounde~ m a stm~hfied or yield the result that giving knowledge (since it is beneficia!) is part of a
prototypical schema of ccrtain arcas of human expcnence. Thts, 1 su~­ general goal of helping others. Thus, in cases in which (2) is true, (1)
gest, is why Coleman and Kay found that lie ~ee~s a protot~pc defiru- translates at least partly as (3) Try to inforrn others.
tion. Basing my analysis on their experimental fmdmgs, 1 mouvate those The rules just proposed constitute the cultural motivation for a folk
findings by relating them to work on discour_se pragmatics and co_nversa- understanding of language as informational. Before going on to a folk
tional postulates. lt is necessary to examme folk understandmgs of theory of knowledge and information, one issue needs clarification: the
knowledge, evidence, and proof; our cultural model o~ languagc ~or at status of these cultural models, or folk theorics. What does it mean to
Ieast of lying) cannot be analyzed independently of behefs about mfor- say that language is assumed to be informational in the "default" case?
mation. 1 hope to show that /ie has a simpler definition than thas been 1do not mean that purcly informational discourse is statistically more com-
thought, in a more complex context; since the cultural-mo~e~ co.ntex~ ~or mon than, or acquisitionally prior to, other kinds of discourse; indeed, it
a definition of lie is independently necessary, our analysts ts stmphfted would be hard to scparate discourse modes cleanly, since one ulterance
overall. may have multiple purposes. However, the informational mode is the
"direct" mode on which indirect speech ls parasitic; aud it may be viewed
as more basic in the sense that all discourse involves the conveyance of
A cultural model of language information (if only about a speaker's intentional state), whereas not all
Is there a simplified "prototypical" speech-act world, as there is a simplified discourse participates in all of the other purposes of language use. Our
marriage history? Allhough such a world has nol been examined in detail, cultural model presentS this "basic" discourse-mode, which is a vehicle
Kay (1983) suggests that the word technically evokes a "folk theory" of for other modes, as being in its pureform the unmarked mode, the norrn.
language use that assumes that cxperts are the arbiters, of corree! ~~rd­ Unlike maxims and conditions, this cultural model docs not constitute
use. Grice's (1975) conversacional maxims, and Searles (1969) febctty- rules of language use, but rather beliefs about what we do when we use
conditions are constraintS on the appropriateness of utteranccs - speakers language. These beliefs in turn rnake generru social rules applicable to the
are assum~d to follow these rules in the default situation. domain of discourse: Grice's maxim of informationality is the manifcsta-
Kay's folk theorics, Grice's maxims, and Searle's ~elicity-c~ndition~ all tion of a general " Help not harm" maxim, in a simplified (folk-model)
describe parts of our cultural understanding of dtscourse-mteracllon. world in which information is always hclpful. Now, on to our cultural
Grice's "Be as informative as necessary," for example, is a maxim of whicb model of information.
speakers are conscious; one can criticize an interlocutor for informati?~al
insufficiency. But informational content is irrelevant to ~ speech acl!vtty
A folk theory of information and evidence
such as jokc-telling. Robín Lakoff's (1973) work on pohteness rules and
Goffman's (1974) work on frame semantics show that conversation often Any truth-conditional semantics assumes that wc can "know" the proposi-
has its primary purposes at the leve! of social interaction; making someone úonal content of "true" statcments; this begs thc vexed qucstion of what
46 EVE E . SWEETSER THE DEFINITION OF LIE 47
knowh:dge is. 1 intend to pass over the philosophers' view of knowlcdge input more than others; and we constantly make (nonlogicaJ) deductions
and instead examine our cultural idea of what counts as knowledge, since based on our observations of correlations in the world. We do not bothcr
this is what underlies our understanding of líes and truths in discourse. to distinguish these generally trustworthy deductions from "fact" except
Clearly, we do not imagine that all our bcliefs can be provenlogically. when observed correlations break down and deductions fail.
Nonethelcss, we consider our beliefs sufficiently justified, and we are not Whatever our rules of practica) everyday inference are like, we trust
really worried that their truth is not known from logical proof (few of them, in the default case. Thus, belief is normally taken as having ade-
"us" speakers know formal Jogic) or personal expcrience. EvaJuation of quate justification, and hence as equivalent to knowledge, which would
evidence is thus frequently an important issuc: "knowledge" is not so much entail truth. Gordon (1974) dcmonstrates the close, complex relationship
a relationship betwecn a "fact" ( = true proposition) and a knower as of belief and knowledge in our cultural understanding; he shows that, in
a socially agreed-on evidential status given by a knower toa proposition. adult as well as child use, factivity of verbs such as know is not fixcd,
Rappaport (1976) demonstrates just how "social" thc difference between especially if the person said to "know" is not thc speaker. A theory of
statement and truth, between beliefand knowledge, rcally is. He observes knowledge as a cultural status given to certain beliefs is more compatible
that a normative standard of truthfulness in informational exchange is with this flexibility than is a theory of knowledge as a link between an
essential to ensurc that our belicf-system (and our social cxistcnce) is not objective fact and a person's mind.
constantly undermincd by distrust of new input. (Actual statistical In our cultural model of knowledge, the default case is thus for belicf
Jikelihood of a random statement's truth is irrelevant to this norm.) He to entail justification and hence truth. Conversely, untruth will entail lack
argues that a central function of liturgy and ritual is to transforma state- of evidence and impossibility of belief. Lct us combine thesc cntailments
ment or belief into accepted, universal truth- that is, in lo something thal with the informational modcl of language. 1 start with a norm-cstablishing
can be unconditionally belicvcd and treatcd as reliable. "meta-maxim":
Rappaport is mainly conccrned with social "facts," not with such
(O) Peoplc normally obey rules (this is the default case).
falsifiable information as "Ed is in Ohio." But lct's remember that
knowledgc has many socially acccptable ("valid") sourccs - and that we Our general coopcrative rule is:
do not in fact tidily separate mcssy socially based knowlcdgc from clean (1) Rule: Try to help, not harm.
falsifiable facts. We know promises can get broken - yet certain ritual
aspects of oaths and promises stiiJ make us treat thcm as extratrustworthy, Combined with a belief such as (2), we can instantiate (1) as a Gricean
maintaining our social norm of truthfulncss. Or, take a modero scholar conversational rule of informativcness, as in (3):
who "knows" Marx's or Adam Smith's economic tcachings - this "knowl- (2) Knowlcdge is beneficial, helpful. (Corollary: Misinformation
edge" may seem toa cynic as faith-based as religious belief, but that does is harmful.)
not prevent a whole community of social scientists from acting on it as (3) Rule: Give knowledge (inform others); do not misinform.
fact. Hard scientific knowledgc and evidence oftcn turn out to be as
Our model of knowledge and informalion gives us the following proof
paradigm-dependent as sociaJ-science argumentation. What is crucial is
of (6) from (4) and (5):
not whether sciemists always have objectively true hypotheses, but that
any society agrees on a range of socially acceptable methods of justify- (4) Beliefs have adequate justification.
ing belief; without such agreement, intellectual cooperation would be (5) Adequately justified beliefs are knowledge ( == are true).
impossible. (6) :. Belicfs are true (are knowledge).
What counts as evidence or authority is thus a cultural question. In
(6) allows us to rcintcrpret our helpfulness-rule (3) yet again:
reply toa college student's scoffs ata medieval philosopher who appealed
to classicaJ authority, 1 once heard a professor ask how the student "knew" (7) Rule: Say what you believe (since belief = knowledge); do
what Walter Cronkite had told him. Many natural languages formally not say what you do not believe (this == misinformation).
mark with evidentiaJ markers the difference between direct and indirect Thc hearer, in this cultural model, is presumed ready to believe thc
(linguistically or Jogically medinted) experience, and/or between various speaker; why refuse help from a speaker who is assumed to be not only
sensory modalities as sources of a statemcnt's information. Sorne priority helpful but also well-informed (having well-justified belicfs)? Putting
or preference seems to be givcn universally to both direct experience together the whole chain of entailmcnts, we reach the startling conclu-
(especially visual) and culturally accepted ("universal") truths. But failing sion that (in the simplified world of our cultural model) the speaker's say-
these best sources of universal truth or personal experience, we trust sorne ing somcthing entails the truth of the thing said:
48 EVE E . SWE ETSER THE DEFINITION OF LIE 49

(a) S said X. Figure 2.1 gives a taxonomy of speech settings; the box on the right
(b) S believes X. (a) plus (7) and the meta-maxim) endoses the idealized informational-discourse world. Lie must be defined
(e) :. X is true. (b) plus (6)) within this restricted world; outside of tbis world, the word lacks applica-
tion. Only within this world can the hearer properly link utterance with
Logically (outside our model), or statistically, this conclusion is rubbish. informativeness, sincerity, and factual trutl1. The feature ( + Truth Value
But as a folk model of Janguage by whicb we al! operate from day today,
Relevant] on the tree indicates that the informational-exchange view of
it makes good sense - in fact, it seems doubtful that we could ever Jive
language is in effect; when truth value is rclevant, knowledge is beneficia}
our lives questioning the truth of every statement presented to us. We ques-
tion truth if we fear that our simplified discourse-world is too far from ACTIONS ( "Help, dc n't htJm")
reality: when our source might be ill-informed (a broken link between bclief

~-'"
and justification), nalve (breaking the entailment between justification/
evidence and truth), or might want to deceive us (invalidating our assump-
tion that folks are out to help, and so wish to inform correctly). Note

1
that even in these cases, the usual cultural model is in effect: W e know ~
our interlocutor expects us to takc what is said asan instance of informa- Ouenlon Ord" ~
tion-giving. But in general, we take people's word.
The next scction examines cases in which we should not take someone's - Trvth Vatue + Truth Value (''lnforrn-'ion •
word; we now look at lying in the simplified discourse-setting established Relevant Renltnt helpful" ..tell

by our cultural understanding of linguistic exchange as informational. 1 (speaker s.upposed the truth"l
1 to góvo koowledge)

Prevarication in a simplified world

Coleman and Kay proposed three components of a prototype-definition
1 flction
of lie:
¡
{ftoty)

l. Speaker believes statement to be false. 1 - Know +-Know

2. Speaker said it with intent to dcceive. (sp.:~aker not fullv (1peaker fully
infonnl!d: belief tnfonned : belltf
3. Thc statemcnt is falsc in fact.
1 ¡tlknowl.ctge) • kr.owlcdgel
Now, in the simplified world we have outlined, any onc of these condi-
tions would entail the others. In particular, if we assume both a folk model
of evidence in whicb a speakcr's be!icf constitutcs evidcnce of truth and
1
J T
(anomatoud
-------' F T
Tellthe
~
Lit

~
Truth
a model of discourse as inforrnational (intending to be believed), then we 1
find that a factually false staternent must be known to be false by the
speaker, and (if made) must be intended to induce (false) belief and thus 1

~
1
to deceive. The reasoning runs as follows: t adequate - adequate
r1idenee or
Premise: X is false. \ jut1iUcation lustifiCJtion

11 -~

\ \
So S did not believe X, since beliefs are true.
Honttsl C.releu
Thcrefore S intended to misinform, since we know that in order MÍ1Uike Mísu.Jce
to inform one says only what one believes.

\ \
Further, assuming that even uninformative speakers do not randomly
discuss arcas in which they have no beliefs (people act purposejully), we
can go beyond "S did not belleve X" to assert "S believed X to be false."
We do not premise the meta-maxim that S is obeying the rules, since S's
obedience to the Cooperative Principie is precisely what we are trying to
prove or disprove.
~ ---------
-- - ---- --
"'-..

Figure 2.1. A taxonomy of speech settings

Social
lie

~
"V/hite "
or juuified
lie
50 EVE E. SWEETSER THE OEFINITlON OF LIE 51

and informing helpful. ( + Know) indicates that our folk theory of informational pan of our simplified cultural model of discourse). Being
knowledgc and cvidcncc is in cffect¡ whcn belicf is justificd and hcncc the first tree-branching above the box enclosing thc simplified world, this
true the speaker can be assumed to have knowledge about what is said. feature is most importan! in speakcrs' judgments as to whethcr wc are
Thus wc can define /ie as a false statement, if we assume the state- in that world (and hcncc whether the tcrm lie applies}. The ncxt tree-
ment oc~urs in a prototypical (informational) speech setúng. This defini- branching, + 1 - Truth Value Relevant, corrcsponds to Coleman and
tion is elegant and would also help explain w.hy native spe~k~r.s ten~ to Kay's "intent to deceive"; a falsehood can only intend to deceive if truth
define líe as a false statement. Not only is thts the first defmtuon gtven value is assumed to be relevant (information = bencficial) - not if we
"out of the blue" by many speakers, but it is (according to Piaget (1932)), are joking or story-telling. This branching is above the + 1- Know branch-
also common for children to pass through a stage in which lie is used to ing and farther from the break between the simplified world and other
denote any false statement. Wimmcr and Pemer's (unpublished data} more worlds - so it is a less importan! feature in a definition thal crucially
recent experimental work shows that children up to age nine class "good depends on that break.
faith" false statements and líes as alike, cven whcn they themselves are Colcman and Kay's lcast important featurc is thc definitional one: fac-
tricked into being thc "good faith" false informcr. Four-year-olds undcr- tual falsity. In the environment of their experiment, which actively
stand sabotage (physical manipulation to obstruct a precondition of an stretched speakers' consideration beyond the prototypical informational
opponcnt's goal) wcll; but five-year-olds are only stan~ng to ~nd~rstand setting, falsehood does not distinguish lies as a unified class. Within the
manipulation of an opponent's bclief-system. Thc soctal m~uvaoons .of simplified world, however, truth val u e criteriaUy distinguishes between
such manipulation entail an understanding of the speech settmg as soctal the two possible kinds of speech act - hence falsehood beco mes the defin-
interaction. Children only cometo differentiate líes from other falsehoods ing characteristic of lie, and native speakers reasonably cite it as such.
as they learn the sociocultural background of speaking and acquire 1he Thomason (1983) (who also tries 10 ground Coleman and Kay's analysis
folk theorics that are a backdrop 10 the more restricled adult use of /ie in the specch setting) adds two more features to the semantic prototype
as a falsc statement made in a certain world.s of lie: "unjustifinbility of belief" and "reprehensiblencss of motive." How-
A fascinating parallel to child usage is found in Gulliver's explanation ever, he himself remarks that unjustified belief in the trulh of X directly
of lying 10 the Houynhnms. His definition, "saying the thing which is not," conflicts with "speaker believes X is false," which he retains; how could
is perfectly comprehensible 10 him, but preves incompre~ensible t~ the both be pan of the meaning of líe? Undcr my analysis, the general maxims
Houynhnms, precisely becausc (as Gulliver says) they have httle expcncnce enjoining us to inform will also condcmn misinformation, even if not
of dccepUon in any arca¡ they lack thc sociocultural background that makcs deliberate. Thus, unjustificd statements will automatically bejudged as like
a falsehood a Iic. Adult English speakers (like Gulliver) have a complex líes in sorne ways (without changing our definition of lie = false statemem
set of possible discoursc-worlds (cf. Figure 2.1); it is nol strange that in in prototypical informative setting). Mere unjustified (sincere) belief does
one seuing ( + Trulh Value Relevan!, - Know} a false statement should not, however, greatly contribute to my actual classijication of cven a fa/se
be called a mistake, whereas in another setting ( + Truth Value Relevan!, statement as a Iic. Furthcrmore, if "unjustificd belief" were part of a defini-
+ Know} a false statement is a lie. . . tion of lie, then even tme, sincere, unjustified statements would have to
Thus, the simple definition of lie as a false statement ts ?atural g~ven be considered líes to sorne degree: not a promising result of an admit-
an understanding of our cultural model of knowledge and dtscourse. The tedly self-contradictory definition of lie. The informationality maxims give
taxonomy of specch settings in Figure 2.1 also rnotivates the arder. of ~ole­ a more general, coherent explanation of any perceived likeness bctween lies
man and Kay's three features. First, it is clear why factual falstty ts the and unjustified statements. We shall scc that Thomason's proposed feature
least importan! feature. Outside of the prototypical (info~mati.onal) specch of reprehensibility also follows from a more general understanding of infor-
environmcnt, falsehood is not particular! y connected wuh lymg (we shall mational exchange and is superfluous to a defmition of lie.
see that /ie's moral status also depends on this setting; for now, suffice Notice how rules and maxims changc formas they change setting: The
it that we cxperiencc a false statement differently when factors like truth- general "Help don't harm" is manifcsted as "Inform others" in the setting
relevance vary}. In a sense, /ie is closer to te/1 the tmth than to joke, al- in which information/truth is the most relevant beneficia! factor. In the
though jokcs are often factually false. domain of politeness, thc same general supermaxim is manifcslcd as R.
Coleman and Kay's most important featurc, the speaker's belief that Lakoff's (1973) politeness rules. This model agrees, 1 think, with our expe-
the statement is false, corresponds to m y + 1- Know branching: Givcn rience: Both information and politeness are considered good and helpful
that a statement is false (anotlzer Coleman/Kay feature), the speaker's cor- (in their contexts), although in fact thc two may connict wben we are un-
rect bclicf in its falsity mere!y constitutes full and correct information (thc sure which setting takes priority.
 EVE E . SWEETSER TH E OEF I NITION OF LIE 53
52
A Jie, then, is a false statement made in a simplified informational- mational setting, such as weakcned truth- vaJue relcvance (litcrary, not in-
exchange setting. All rules enjoining veracity are in effect, a nd the speaker formative goaJs) or less complete control of facts by thc speaker, can make
is a fully knowledgeable imparter of information to a credulous hearer. the difference between our judging a falsehood as a lie (within the
Lie has a simple definition within a matrix of cultural models that are simplified informational world) oras something else (in sorne other world),
independently neces.sary. The prototype seems to be in the context, rather such as a tall tale.
than in the definition itself. Speakers havc difficulty judging whether an Mistakes are cases in which, without speakers' knowledge, the normal
action is a lie when they are not sure the action's setting sufficiently matches chain of entailmcnt from belief to truth breaks down. Both speaker and
the prototypicaJ setting specified by the cultural mod:t of informati~nal heare r think they are in the simplified world delineated by cultural models
exchange.6 The next section fits a larger sector of Engltsh ~ocabulary mto of knowledge and evidence, but there is an unknown dcviation. For an
the cultural model we have outlined; 1 then go on to mouvate our moral honest mistake, in parúcular, the entailment betwcen belief and evidence
condemnation of lying in terms of our cultural models as well. does hold: The speaker has nonnaJiy s ufficient reason to believe what was
said. Carelessness is charged if the broken entailment is between belief
and evidcnce - the speaker should ha ve rcalized the cvidence was insuffi-
Less simplified worlds, less simple words cient, but failed to. Speakers are responsible for evaluating evidence, so
English has words for false nonJies, or palliatcd/justified lies. These words we blame irresponsibility where we would not blamc an honest mistake.
mark deviations from the simplified world of the cultural model; thus, ln cither case, however, we assumc that thc rules ought to hold: Mistake
examining the deviaúons may elucidate the mo del. Common terms include marks a disruption o f our simplified informational world's assumptions,
white lie, sociallie, exaggeration, oversimp/ification, tal/ tale, fiction, fib, rather than an agreed-on suspension (in favor of othcr goals), as in the
and (honest or careless) mistake, sorne of which appear in Figure 2. 1. case of joke. Lie, on the other hand, denotes a wrong moral choice, with
first as stressed in the previous section, a lie is not committed if truth no disruption or suspension of the informational model.
¡5 irrele~ant. Thusjokes, kidding, and leg-pullings, which exist in a world As further indication that speech acts are subcases of actions (rathcr
where humor rathcr than information is the basic goal, are outside the than sorne separate, parallel category), note that the same word mistake
infonnational model and cannot be considered lies. Of course, every denotes both an unintentional falsehood and a wrong turn ta ken, or a
culture al so has a mo del for humor, and humorous discourse (like aJI typo. ldeally, we shouJd be able to justify any act, speech or otherwisc;
speech) uses sorne aspects of the informational model. Whcn wc cannot the graver the conseQuences, thc higher the standards for justification.
decide which model predominates in a given situation, we ask the ~om­ But blameless wrong choices do occur; and if we did our bcst with avallable
mon (and intelligible) question, "Ho w serious was that remar k?" Sen~us­ information and resources, unimentionaJ harm can be forgiven . The cate-
ness characterizcs contexts, not statements; the same remar k may be senous gory mistake is a recognition of human frailty as an allowable out.
or not, depending on context. Since interlocutors constantly negotiate con- In exaggerations, oversimplifications, understatements, and other dis-
tcxt (including the prcdominancc of informational or humorous goaJs), tortions, the informational-cxchange rules are more or less consciously
one may ask about a statement's seriousness, meaning the speaker's percep- bent, rather than suspendedor disrupted. Such cases do not strictly follow
tion of its micro-discourse context. . . thc dictates o f our cultural model; we feel we are being less informational
Tal/tales, fiction, andfantasy, when not referring to hterature, pa.lhate Oess truthful) than we might be, hence less helpful. But distortions are
falschoods by looking at them as literary, rather than as prototyptc~ly not nccessarily in direct opposition to truth; they may indicatc a subjec-
informatiooaJ. The discourse in quesúon is lookcd at more as a story (wtth tivc personal rcaction better than thc strict truth could, and hence be
a goal of artistic entertainment) than as facts with relevant truth vaJues. truthful at another leve!. Or, it may be more informational for an expert
Grandpa's tal/ tales of fifty-foot snowfalls in his childhood are f~n and to oversimplify than to fail totaJ1y to commul\Í.cate with a nonexpert. Many
harmless. Similar claims in a history book, however, would be nustakes, such distortions are indisputably literally falsc. Whether we judge them
to say the least. Tal! tales of huge fish l caught are li~s if we are still o_n as lics depends o n (1) whether the setting is prototypically informational
the fishing trip and 1 convincc you there is fish for dmner wben there ts and (2) if so, whether they advance or obstruct the informational goals
not. 1 personal! y only use fantasy and fiction to refer to literature (or to of interaction.
interna!, unspokcn fantasizing). Whenfantasy refers toa false statement, Wlzite lies and socia/lies are gcnerally like lies, but they occur in set-
howevcr, it seems not only to mean a more artistic story than t~e truth, tings in which information might harm rather than help. Thcy are still
but also to include an element of seif-deccption that further p~lltat~s the called lies: even nonreprehcnsible, deliberate misinformation counts as a
offense o f deceiving others. Any departure from the prototyptcal m for- Iic. In these cases, the entailments of speaker's knowledge, evidence, and
54 EVE E. SWEETSER THE DEFINITION OF LIE 55

intent lo be believed (seriousness) Slill hold; likewise the supermaxim ''Help more or less a líe because of reprehensible morives on the speaker's part
don'l harm" holds; bul the usual helpfulness of lruth cannot be assumed. (consider my Gestapo example as a case of a real líe with good motives),
For a social lie, the politeness maxims have superseded the injunction decide that such motives are typical rather than prolotypical of lying. That
to lruthfulness. Truth is seen as more harmful to the social situation than is, lies tend in the real world to be selfishly motivated, j ustas real surgeons
minor misinformation would be. In the case of white lies, truth might harm currently tend to be maJe; but one cannot claim that maleness is in any
in some other, sometimes more direct, way: Sorne people would call it way part of the meaning of surgeon.
a while Iie to tell a dying person whatever he or shc nceds to hcar to die Placed in the framework of cultural models of discourse and informa-
in peace. Sorne speakers would also call a (less altruistic) lie told in self- tion, the variable reprehensibility of lies follows naturally. To the extent
defense a white lie if it helped them and hurt nobody else. As with polite- that information really is beneficia! at a higher leve), and false informa-
ness, self-defense is clearly only supposed to be allowed to supersede tion harmful, a lie will harm. General social judgements will condcmn
the informalional mode if the consequences of the resulting deception deliberate harmful actions.
are small. The compounds white lie and social lie show in their two Thomason (1983) disagrees that lies are typically reprehensibly moti-
elements the conflicting worlds in which the actions take place (it is valed; he suggests that social lies are the most common sort of líe a nd
a lie as an informational utterance. but it is also a social utterance). are nonreprehensible. 1 differ wi1h him; social líes are rarely altruistic,
Figure 2.1 puts them under more than one heading lo show tbis dual though their element of selfishness may not be deeply harmful; and their
categorizalion.7 statistical prcdominance is unprovable, as a valid survey is surely impossible
There are lies which most people would think justified by sorne higher in this domain. Coleman and Kay correctly renect a folk understanding
good achic:ved but which would not be called white lies, since their infor- 1hat deceit usuall y profits the deceiver, to the listener's detriment.
mational consequences are 100 major (however moral) for us to diminish Thomason's wish to include reprchensibility in the prototype of lie shows
their status as lies. 1 would think it moral to lie to the Gestapo about the that he s harcs this folk belief in a deep connection between deceit and
location of a Jew, but 1 would call that an unqualified Iic. The informa- harmfulness.
tional paradigm is fully, even saliently, in effect in this instance- it is only This deep judgment of falsehoods as inherent/y harmful goes beyond
that we feel our uncooperativeness to be justified. what we can so far predict from cultural models examined; our informa-
Lasl and least, afib is a small or inconsequential lie, and thus a palliated tional-exchange model would ask us to condcmn falsehood only when,
offensc, sincc the seriousness of an o ffense of lying is a function of its harm~ in facl, truth is beneficia[ and misin formation harmful, so that 1he
fui consequences. However, a fib is nonetheless an offense (though minor) simplified world is in effect. 1 now turn 10 an examination of the cultural
in that it is considered to have at most only a selfish and unimponant reason links bctwcen information and power , in order to explain wby a stigma
for overriding the usual motivations for veracity. of immorality attaches to even well-intentioned prevarication. Let us first
This brings us to the qucstion of the importance of a falsehood or a cxamjne what we do in making an "ordinary" informational statemcnt,
deception. As Coleman and Kay observe, we can only judge major versus true or false.
minor deviations from the truth in terms of human consequences. They R. Lakoff's (1973) Rules of Politeness, now recognized as a neccssary
contrast an error in the millions column of a city's population (a decep- part of our understanding of speech acts, are:
tion) with an error in the oncs column (no deccption, because it has no
l . Don't impose. (Formality)
serious consequences). It is clearly only felt allowable to override the truth-
2. Give options. (Hesitancy)
is-beneficial maxim when the truth-violation could have no negative con-
3. Make interlocutor fcel good; be friendly. (Equa/ity/Camaraderie)
sequences as serious as the negative results of truthfulness. A social lie cannot
be justified as polite (hence helpful) if it grave! y and harmfully misinforms. Lakoff says (2) explains why a direct command is less polire than an in-
When truth is more important than politeness, the informational mode can- direct command with the surface form of a request or o f a query abour
not be overridden. This mere! y repeats that our judgement of a lie depends the hearer's willingness or ability 10 do the task. lndirect forms givc the
on the extent to which the relevant cultural models are in effect. hearer options besides obedience or disobedience; the hearer can negatively
answcr a query a bout ability without having to rcfusc compliance dircctly.
Ahernatively, indirectncss allows compliance without implicit acceptance
Knowledge as power: the morality of lying
of lhe felicity-conditions of a command and recognition of the speakcr's
The cuhural models relevanr to lying also help explain the generally ac- authority. Hedged commands avoid assuming ungranted authority ovcr
cepted reprehensibility of lies. Coleman and Kay, noting that a lie is no an add ressee. Without details of the motivation, Lakoff also says that
 EVE E. SWEETSER THE DEPINITION OP LIE 57
56
the same factors make it more polite to qualify assertions with "1 guess" the normal degree of responsibility for a statement's truth by qualifying
or "sorta." This seems a puzzle at first: Why should it be more polite to its evidentia.l status. Unqualifícd statements presumably take on a default
guess than to assen, orto make a hedged assertion rather than an unhedged leve! of rcsponsibility, varying with context.
one? Statements bave so many purposes that the issue is messier than for However, evidentiality-hedges have another function besides the meta-
commands, but tbe answer (as Lakoff at least implicitly noticed) is that linguistic evaluation usage just described; they also function as pragmatic
a statement does something to the hearer, just like other speech acts. lt deference-markers. However su re a student may be of one of the follow-
pushes at the hearer's belief-system. An informative speaker requires a ing assertions, he or she might have social motivation to mark uncertainty
hearer ready and willing to believe, or information cannot be imparted. with an evidentiality-hcdge:
This cooperative hearer grants the speaker a good deal of power to push
But, Professor Murray, asfar as 1 can te/1, this paraUels Andrcws'
around certain aspects of bis or her belief system.' example, which suggests another interpretation.
English reOects the equation of knowledge with power, in the uses of
a group of hedges that mark the evidential status of statements. Sorne Professor Jones, if /'m not mistaken, haven't Smith's recent results
examples of evidentiality-hedges are: to the best of my knowledge; so far made the Atomic Charm hypothesis look dubious?
as 1 know; if /'m not mistaken; as far as 1 can te/1; for al/ 1 know; as 1
When social authority is low, the right to push people's belief systems is
understand it; my best guess is; speaking conservatively; at a conservative
correspondingly low. Especially if our hearer may be unwilling to listen
estímate; to put it mildly; beyond question.
Thc literal use of these hedges is to limit the speaker's normal rcsponsi- and change opinions, we havc to be socially careful; we have no more
bility for thc truth of assertions. An assertion has the precondition (Scarle authority to command belief changes than any other action against the
1969) that the speaker be able to provide cvidence for its truth. Or, in wiU of our interlocutor.
terms of our cultural models of information and cvidence, in an informa- Evidentiality-hedges thus hedge both kinds of authority that underlie
tional setting a hearer knows that a coopcrativc speaker will only state an assertion: informational authority (evidence) and social authority (we
justified belicfs. However, even reliablc-looking evidence can turn out to cannot as readily command belief-systems of people higher on the social
be insufficient. Evidentiality-hedges allow the hearer access to the evi- scale). This is a natural pairing, considering our understanding of asser-
dence-evaluation and thus transfer sorne of the speaker's evaluative respon· tion as manipulation of bclief systems. In a prototypical informationnl
sibility to the hearer. They avoid potcntial charges of carelessness or exchange, the hearer is as ignorant and credulous as the speaker is knowl-
irrcsponsibility by not allowing the hearcr to over- or undervalue the evi- edgeable and ready to inform. Who has the upper hand in sucb an ex-
dence supporúng the hedged assertion. (Cf. Baker 1975 on sorne related change - the knowing and manipulativc speaker, or the ignoram and
hedges that signal and excuse potential discourse violations.) passive Jearner? Teaching (a rclatively onc-way exchange, atleast in early
G. Lakoff points out (personal communication) that responsibility- stages) has aspects of authority cven without a surrounding institutional
transfer goes even further. Not only can we qualify a statement's eviden- power-structure. Toa lesser degree, any assertion has the same inherent
tial status, but we can also evade personal responsibility for the original power structure.
In further support o f this analysis, note that a person with both kinds
(prequalification) statement. For example:
of authority can lay aside either kind with an appropriate evidentiality-
to thc best of our current knowledge hedge. A professor who wants to get a point out of a studem rather than
to the extent to which this phenomenon is understood at al/ giving the answer may thus !ay aside both aspects of authority, in a state-
ment likc: -
so far as can be judged from work to date
according to the current consensus in the jield But as 1 understand it, semantics is the study of meaning- so how
docs it strongly depend on spelling, Mr. Smith?
This last set of hedges makes criticism or disagreemcnt difficult; whereas
if the speaker had simply evidentially qualifíed his or her personal evalua- Too many such hedges from the professor would sound sarcastic, sincc
tion, the hearer cou\d easily disagrce (though not accuse the speaker of it is insincere to deny the existencc of one's power position while leaving
irresponsibility or prevarication). At the opposite end of the spectrum, its broadcr social presence unchanged.
hcdges such as speaking conservatively commit a speaker to an assertion's As further evidcncc that spcakers link asscrtion with (a) request for
high evidential status (another example is a/1 the evidence points lo the belief and (b) assumption of an authority position, consider the follow-
conclusion that). Evidentiality-hedges, then, allow thc speaker to modify ing hedges:
58 EVE E. SWEETSER THE DEFJN ITION OF LIE 59

(Piease) believe me: ... tional authority is warranted OR (2) even greater social authority than
1 don'/ ask anyone lo believe lhis, but . . . a command; the hearer is expected not only to obey, but also to deduce
and meet the speaker's wishes before they are stated (the hearer does not
1 can'/ expecl you lo believe me, but .. . °
seem to mind the open window). 1 For me, the politeness-contrast re-
These hedges mark unreasonable belief-requests, tacitly assuming that an verses (as expected) if "The window's open" is said courteously, toa per-
ordinary belief-request is justa matter of course. I can'! expect you lo son who somehow (mental absorption? a physical barrier?) j ust has not
believe me needs to be stated, even though our normal right to such an noticed but might reasonably share the speaker's concern. These examples
expectation passes unnoticed and unstated . demonstrate the complex interplay between informational and social au-
Phrases like the strength of an assertion, or the authority for a state- thority in determining politeness.
ment, are not random. Both social and informational authority structure From the preceding discussion, lying emerges as serious authority-
our discourse world, and the strength of an assertion depends on both . abuse. Authority relations structure the prototypical informational ex-
lf either kind of authority is extremely strong, it may overcome opposi- change, the setting in which lie is defined. As we get further from the
tion from the other: An undergraduate who is very sure of a fact may simplified world in which the credulous hearer depends on the speaker
correcta department chair, anda dean may feel freer than a student to for sorne crucial information, truth becomes less relevant and falsehood
speculate, having more social protection from contradiction. lcss reprehensible. In the simplified world, however, (barring major reversa!
Thus, our cultural model of information as power motivates eviden- of social authority and morality judgments, as in the Gestapo example),
tiality's relationships with politeness and authority. lncidentally, Grice's falsehood constitutes a deliberate use of authority to harm someone in
( 1975) maxims are often cited as barring assertions that are obvious or a wcaker, dependent informationa1 position. We thus naturally judge it
well known to tbe hearer because tbey are useless and uninformative. How- as immoral, barring exceptional extenuating circumstances.
ever, 1 have not scen it overtly said that obvious statements are also often As salient examples of our view of lying as authority-abuse, Jet me cite
insulling. Their rudeness cannot be dcduced from their uninformativeness the anger of patients lied to by doctors, or of children systematically Jied
but follows dire.ctly from viewing them as unwarranted assumptions of to by adults (e.g., about sex). Doctors in particular derive much of their
9
informational autbority ("1 know better than you"). This view may help authority from large amounts of knowledge that is not otherwise acces-
explain the Coleman example (P. Kay, personal communication) "Crete sible to patients. By refusing information or misinforming, they can con-
is sort of an island," where sor/ of appears to hedge neither the choice , trol importan! decisions for patients. To a lesser dcgree, any possessor
of the word is/and nor the precision of the truth-value, but the act of as- of information can influence or control less knowledgeable hearers. To
serting is weakened to avoid rudeness. the extent that we feel people should control themselves, lying is immoral
Conversely, Jef Verschueren (personal communication) points ou~ to because it undermines the potential for self-determination. " This deep
me that the idea of informational authority gives added motivation (bes1des idcntification of lying with power abuse may explain why for sorne peo-
Lakoff's rules) for seeing questions about ability or willingness as politer ple all lies retain sorne reprehensibility, however good the motive.
than direct commands. Qucstion form has the inherent courtesy of giving
the addressee a presumed informational authority. It is no huge politeness Deception and lying
to assume an individual is the best authority on his or her own wishes
and abilities. The contrary assumption, however, is ipso Jacto particu- Líes are only a subclass of deception. Any deception, in that it induces
larly counter to the rules of politeness, unless either camaraderie or un usual false beliefs in a credulous hearer, is a culpable abuse of informational
social authority overrides politeness. A direct command thus indicates pre- authority and naturally liable to the samc moral charges leveled at a Jie.
sumed unconcern for whether the addressee has opinions, let alone what But oddly enough, speakers often feel less immoral if they manage to
they are - and in a domain in which that person is the evident authority deceive rather than to lie straight out. Victims conversely feel that such
(i.e., his or her own interna! state). a deception is a dirtier trick; they cannot complain of being Jíed to and
Verschueren also drew m y attention to the contrast between an indirect resent the dcceiver's legal loophole.
but less polite "The window's open" (in a rude tone, to hearer who sees T here thus seems to be a further folk belief that literal truth and real
the window) and a direct but more polite request or command "(Picase) truth (honest information-transmission) are prototypically connected. A
close the window." Here 1 feel, the chosen mode of indirectness is more literally true statement thus retaíns vestigiallegality (if not morality), even
insulting than a direct command - the statement implies either (1) tbat if it misleads, whereas a delibera te factually false statement retains sorne
the hearer is so unaware of the obvious that the assumption of informa- stigma of reprehensibility, even with strong moral justification. Folklore
60 EVE E . SW EE TSER THE DEFINITION OF LIE 61
gives magical power to literal truth, and a common folk theory is that linguistic communities also share the accompanying moral judgernents of
law also emphasizes literal truth rather than informativeness (1 do not know lying, probably due to shared undcrstanding of power structures and in-
about modem perjury Iaws). Sorne people would find lying to the Gestapo formational cxchange. However, a first glance at more distan! cultures
immoral; yet most of them would think it laudable to mislead villains, shows a startling degree of surface variance asto the morality of rnisleading
saving an innocent victim. In any case, complete dissociation between or lying. Ochs Keenan (1976) discusses the frequency (and acceptability)
literal and "real" truth, or between the latter and morality, is regarded of vague or misleading answers to questions in a small Malagasy-speaking
as highly atypical. community. Gilsenan (1976) states that successfullying is a major positive
A common way to mislead is to imply, but not overtly state, the false status- source formales in a Lebancse Arabic-speaking community. In what
proposition to be communicated. The oven statement and the false prop- respects do these groups differ from English speakers?
osition are often linked by Gricean conversational implicature; the utter- My answer is that, on examination, these cultures differ from ours much
ance is irrelevant or insufficient in context, unless the hearer also assumcs less than the isolated statements above might indicate. At least, the dif-
the unspoken falsehood. In such cases, the speaker could without self- ferences are not in their understanding of informational exchange,
contradiction go on to cancel the deceitful implicature. Taking a case from evidence, or abuse of informational power.
Coleman and Kay: "Mary, have you seen Valentino lately?" Mary: "Valen- Ochs Kcenan's Malagasy community, while agreeing with English
tino's been sic k with mononucleosis all week." Ma ry could go on, "But speakers that information-giving is cooperative and useful, has a different
!'ve visited him twice." Part of peoplc's disagreement about thc morality idea of when a hearer has a right to such cooperation. Europeans or Amer-
of misleading (and about whether it constitutes lying) m ay be genuine dis- icans might think of thcir own contras! between "free goods" (any stranger
agreement about the degree to which a conversational implicature consti- gets a reply to "What time is it? '') and other facts (e.g., one's age, or miel-
tutes a "statcment" and hence makes the speaker responsible for having die name) that need a reason to be told. Malagasy speakers place an even
said it. As Thomason says, sorne speakers are so su re the implicaturc was higher power-value on information than do English speakers (news is rare
present that they include it in a restatement: "Mary said No, Valentino in small communities) and naturally hoard prccious and powerful knowl-
had been sick." edge; questioners cannot expect as broad a spectrum of free goods in such
The plot thickens as the implicatures become more closely bound to a society, and day-to-day informational demands have less right to ex-
thc linguistic form. Such implicatures seem tome to be closer to statements pect compliance. Malagasy speakers are not uncooperative when refus-
than Mary's implicature about Valentino. Thus, 1 would predicr that an ing information could scriously harm (e.g., if asked "Where's the doctor?"
utterance such as "Some of my students cut class," (used when not one by an injured person). Our classic informational-exchange setting is just
showed up) would imprcss speakers as closer to a prototypical lie than not in place as often as in an English-speaking community; since Malagasy
Mary's statement. speakers all know this, their equivocations do not manipulate unsuspect-
An evcn more difficult case is that of presupposed falsehoods. How ing addressees. The Malagasy community shares basic cultural models of
close to lics are statements such as "He's only a sophomore, but he got information and truth with English speakers, but evokcs them under dif-
into that course," used of a student at a two-year college where sopho- ferent circumstanccs.
mores are themost privileged students, and said to deceive the hearer about We might note hcrc that lying to cnemies is often culturally accepted.
the nature of the college or the course? 1 personally rate these examples Many English speakers think such líes less immoral than líes to trusting
high. 1 hope in the future to investigate what constitutes "stating," as well friends, who are "owed" more sincerity (Coleman and Kay cite speakers
as what constitutes Jying. Our cultural model of representation is essen- who, extending this scale, said Mary did. not "owe" John the truth about
tial to our understanding of misreprese.ntation. Valentino, as they were not engaged). I n sorne cultures, lying may be for-
bidden primarily within the group; but such a culture does not lack our
Cross-cullural paral/els judgment of lies as harmful. Rather, their rule about who should not be
harmed is different.
Anthropologists interested in c ultural models, or linguists interested in Gilsenan's Lebanese village is a n even more complex case. He states
culturally framed semantics, now ask "How universal or culture-bound that this community thinks lying immoral, probably for the same reasons
are the cultural models we ha ve just examined?" 1 ha ve used English data wc do. Community members caught lying lose status and honor. How-
(like Coleman and Kay); studies of French (Piaget) and German (Wimmer cvcr, certain rcstricted kinds of undctected líes told by adult males can
and Perner, above) child Ja nguage agree with each other and are highly be extremely status-productive.
compatible with my proposed analysis of the English ver b lie. These First, verbal self-presentation is highly competitive for Lebanese men,
 - .- -· ..... --·... _... THE DEFINITION OF LIE 63

so false (or unfalsifiable) boasts are profitable, though. ~etec~ion cau~es more complex than this simple definition indicates. Definitions of moral! y,
corresponding status-loss. Conventional verbal competltiOn ~·ves nonm- informationally, or otherwise deviant speech acts follow readily from a
formational aspects to Lebancse boasts (though not as formahzed as, e.g., definition of a simplified "default" speech world. The cultural models in
Turkish or urban black American, boys' boasting or insults). English question not only underlie a whole sector of our vocabulary but also moti-
speaker~ might lie competitively in other arcas, and less conventionaUy; vate our social and moral judgments in these arcas; they further appear
but the Lebanese view of lying is not in serious conflict with our own. to have strong shared elements cross-culturally.
The second way a Lebancse man can gain status by lying is to lead Cultural models underlying linguistic systems are a fairly new arca of
another man "up the garden path" and subsequently reveal the deccption. analysis, though a few people were ahead of the rest of us (Becker 1975
He must avoid dctection, or it may be difficult to prove he did not mean is a good example). However, collaboration among linguists, anthropol-
to deccive permancntly. A "garden path" is crucially not real lying, since ogists, and other social scientists in this arca looks increasingly fruitful.
it achieves its goal only by eventual truth-revelation. Thus, such deccp- M y own preference for this approach stems from both its intuitive plausi-
tions do not show a differcnt idea of lying from ours; but why do these bility (ethnographers, if not grammarians, have long known that word-
play-líes give status? . . meanings are interrelated with cultural models) and its explanatjon of a
Gilsenan explains that discernment is a maJor source of prcsuge for long-term paradox facing semantic analysts. Word-meaning has orderly
Lebanese men: A reputation for telling truth from falsehood is valucd aspects that make us feel that it ought to be simply formalizable; yet we
especially in religious leaders, but nlso in any adult mate. He tells of a all know from biu er experience how readily tbe complexities of meaning
visiting religious leader who upstaged the village religious leader (a man elude reductionistic formal analysis. lf the analyst's intuitive feeling that
with a long-built reputation for disccrnment, even omniscience}. A village definitions are simple is right, then perhaps much of the fuzziness and
man, rcsenting the intruder, perpetrated and then publicly revealed a suc- complexity lies in the context of meaning, rather than in the meaning itself.
cessful minor hoax on him; he left, discredited. Lebanese "garden-path" A beuer understanding of cultural models (aided by research such as that
líes are usually less importan!, but do cause real status - gain or loss - representcd in this volume) is important to lexical semantics: Words do
unlike American April-fools or leg-pulling. not mean in a vacuum, any more than people do.
Lcbanese society evidently has conventionalized competitive uses of in- This paper )caves many unresolved problems. It is insufficicnt to dis-
formational power; men overtly gain power by forcing false beliefs on cuss one cultural model or folk theory of speech (here, our default model
others or by seeing through false claims (exposing the auth~r as no~au­ of literal discourse as informational) as if it were largely independent of
thoritative dishonorable or simply unsuccessful at one-uppmg). Senous aU the othcr models relcvam to verbal interaction. Our folk understand-
use of this ~ower by lyin~ would be immoral, but one can conventionally ing of knowledge also needs more investigation. On the linguistic front,
display power without using it - as a martial arts victo~ does not ~ill but in which cases can we expect the fuzzincss of fuzzy semanlics to be ulti-
shows that he has overcome his opponent and could kili. A mart1al ans mately locatable in the sociophysical world (or in our perception of it),
victor's status need not ind.i catc corresponding cultural a pproval of ac- or in the fit between the world and a cultural model; and in which cases,
tual killing or assault; nor should status given by "garden paths" be taken if any, can we expect inherently fuzzy semantics? This last queslion can
as indicating general social approval of lying. be answered only as we learn more about the relalionship between linguistic
Very different cultures emerge from this discussion as possessing sa- and social (even metaphorical) categorization. Just now, 1 must be con-
liently similar understandings both of lying and of the general power and tent with showing that a simpler semantics o f lie follows from an analysis
morality dimensions of informational exchange. This similarity presum- of the cultural models releva nt to prevarication.
ably stems from universal aspects of human communication. Where
cultures differ appears to be in delimitation of basic "informational ex-
change" settings and in conventionaluse of the relevant power paramet~rs. Notes
Folk models of knowledgc and informativeness (and the correspondmg l. Only members of tbe Berkeley linguislic community will understand how much
semantic domains) may universally involve strong shared elements. 1his work owes to thcir ideas and support. However, my intellectunl dcbt to
my advisors, Charles Fillmore and George Lakoff, should be evidem. Linda
Coleman nnd Paul Kay, original inspirers of this pro)cct, were patient and
Conclusions intelligem critics tlrroughout. 1 have also benefited from the insightful com-
ments of Susao Ervin-Tripp, Orin Gensler, David Gordon, John Gumperz,
A tic is simply a false statement - but cultural models of information, Doro1hy Holland, Mark Johnson, Naomi Quino, John Searle, Neil Thomasoo,
discourse, and power supply a rich context that makes the use of /ie much Jef Verschueren, Jeanne Van Oosten, and the participnnts in tbe Princeton
64 EVE E. SWEETSER THE DEFl NITION OF LIE 65

Confcrencc on Folk Models. An earlier version o f the paper was presented tion. One case shc analyzes is that of a woman who was the only likely kidney
in the symposium Folk Theories in Everyday Cognition, organized by Holland donor for her daughter and overtly willing. Perceiving severe repressed fears
and Quinn for the 80th Annual Meeting of the American Anthropological in her, doctors falsely told hcr that she was not physically compatible enough
Association, 1981. . with her daughter to be a good donor. This dec~ption robbed her of the chance
2. The term folk theory, which 1 originally used throughout, emphames the to confront her fears and make her own decision about giving the kidney.
nonexpert status o f such a theory or model; cultural model, which l am now Bok also notes that deception is less frightening if we ourselves have authorized
adopting, stresses the fact that our cultural framework ~odel~ the world for the deceivers and are awarc of their tactics. Unmarked traffic control cars
us. 1 have retained the wordfolk in contexts where 1 find 11 pan1cu1arly useful. voted into use by the community are 1css thrcatening than if the police use
3. For a recent and complete survey of work on linguistic catcgorization, see them without citizens' input.
G. Lakoff (in press). .
4. Coleman and Kay presemed subjects with a series of shon fictional scenano_s,
asking the subjeclS to judge in each case ( 1) whcth.er a lie had been. t~ld m References
the interac1ion described and (2) how sure thc sub¡cct fel1 about th1s ¡udg-
mem. The actions described in thc scenarios varied indepcndcn1ly wi1h respcc1 Baker, C.
to deceptiveness, factual falsity of statements made, and speaker's be1ief of 1975. This is justa first approximation but .... In Papers from the Eleventh
1he contcnt of thc statements. Regional Meeting of thc C hicago Linguistic Society. Chicago: Chicago
5. Susan Ervin-Tripp has suggested to me that young childrcn are simp1y Linguistics Society. Pp. 37-47.
"behaviorists ," judgingacts by result, not by in1ent. Beforechildren can sta1e Becker, A.
thelr imentions, they are bound to get rewarded and punished behavioris1ically. 1975. A linguistic image of nature: The Burmcse numerativc classifier system.
Four- 10 nine-year-olds are certainly not insensi1ivc to intcntions but may re- Linguistics 1975 (165):109-12 1.
main bchaviorists cnough to class lies with other false statcments. Bok, S.
6. Paul Kay has brought 1o my attention a playful usage 1hat seems odd in 1hc 1979. Lying: A Moral Choice in Public and Prívate Life. New York: Pantheon
context of either a feature ora prototypc: ana1ysis of lie: " Do you know, 1 Books.
though1 1 1old 1he truth the other day, ~~~~ it turns ~ur 1 lied t~ you: l'm Coleman, L. and P. l<ay
so sorry." This usagc seems 10 me paras1hcal on scnous usagc m tha1 1he 1981. Prototype semantics: The English verb ' lie.' Lnnguagc 57(1):26- 44.
speaker jokingly attributcs to a past spcech act his or her current mental Fauconnier, G.
knowledgc-space (in Fauconnier's [1985) sense of mental space). Since past 1985. Mental Spaccs. Cambridge, Mass.: M. I.T. Press.
acts are not actually judged in the light of subsequently gaincd knowlcdge, Fillmore, C. J.
we find this amusing. 1977. Topics in lexical semantics. In Current lssues in Linguistic Theory, R.
7. Lakoff (in press) comments that sociallie and simila r collocations pose prob- Co1e, ed. Bloomington: University of Indiana Press. Pp. 76-138.
lcms for the theory of complex categories. A prototypical social lie is not Forman, D.
necessarily a prototypica1 lie. Without proposing a new theory of compl_ex n.d. Informing, rc:minding, and displaying. Unpublished manuscript, Depan-
ca1egories, 1 fcel it i.s clear !hat sociallie is notan intcrsection of the ca legones ment of Linguistics, University of California, Berkeley.
/ie and social act. Rather. it is viewed simultancously (and perhaps somewhat Gilsenan, M.
contradictorily) as a member of two categories that we do not usual!y under- 1976. Lying, honor, and contradiction. In Transaction and Meaning: Direc-
stand as interacting at all. tions in the Anthropology of Exchange and Symbolic Behavior, A.S.A. Essays
8. Social rights and responsibilities are reciprocally arranged: lf the ~pea_ker has in Social Anthropology, Vol. 1, B. Kapferer, ed. Philadelphia: lnstitute for
the right (authority) to say X, then the Hearer has a duty to. behev~ 11. If H the Study of Human lssues. Pp. 191- 219.
has a special right to hear (to know) X, beyond the general n ght to mforma- Goffman, E.
tion then S has a correspondingly more imponant duty to tcll X to H. 1974. Frame Analysis: An Essay in the Organization of Experience. New York:
9. Paui Ka y has suggested 10 me that thc rudeness of telling someonc what they Harper and Row, Publishers.
already know is bcst compared to the rudeness o f giving an unnecessary or Gordon, D. P. _
redundan! gift. However, such gifts are only rudc i/ they imply an unwar- 1974. A developmenta1 study of the sema ntics oi fac1ivi1y in the verbs 'know,'
ranted power-assumption. lf 1 give you a paperback you ow~ a copy o~, l'm 'tbink,' and 'remc:mber .' Ph.D. dissenation, University of Michigan. (Published
only rude if 1 thereby (unjustifiably) purported to extcnd your h1erary hon~ons; as Natural Languagc S1udics, No. 15. Ann Arbor: University of Michigan
bu1 if 1 pay for your bus ticket (which you are presumed capablc of buymg), Phonc1ics Laboratory.)
then l'm rude unlcss you asked for help with changc. All valunble resources, Grice, H. P .
like information, confer power on their owners. 1975. Logic and conversation. In Syntax and Semantícs, Vol. 3, Speech Acts,
10. Forman (n.d.), in a (somewhat astonishingly) still unpublished papcr, " ln- P. Cole and J . Morgan, eds. New York: Academic Prcss. Pp. 113-127.
forming, Reminding, and Displaying," clucidatcs thc in~orma.tional uses of Kay , P.
apparcn1ly noninformativc statemcnts; he would categonzc tl11s asan exam- 1983. Linguistic competcnce and folk theorics of Janguage: Two English hedges.
ple of informative reminding. . . . . . In Proceedings of the Ninth Annual Meeting of 1he Berkeley Linguistics So-
11. Bok (1979) provides a treatmcnt of the soc•al•ssues m volved m lymg and dccep- ciety. Bcrkeley: Universi1y of California. Pp. 128-137.
 66
EVE E . SWEETSER 3
Linguistic competence and jo/k theories
Kay, P. and C. McDaniel .
1978. The Linguistic Significance of the Meamngs o
f Basíc Color Tcrms. oj language
Language 54(3):610-646. T\VO ENGLISH HEDGES 1
Lakoff, G. es· A stud in meaning criteria and the logic of f~zzy ~o.ncept~- In
19~~ Hr~f~o~ the Ei:hth Regional Meeting of the Ch!cago L~ngulsllc. So~lcty~ Paul Kay
C~go. Pp. 183-228. (Reprinted in Joumal of Ph1losoph•ca1 Log1c 2.458
sos. 1973.)
La1~0~~~~.' Womcn, Fire, and Dangerous ~hings: What Catcgorics Reveal about
the Mind. Chicago: University of Chtcago Press.
Lakoff. R. . of litencss· Or minding your P's and Q's. In Papers. from
19~~~ 1~~~~~~gio: Meeti~g of the Chicago Linguistic Society. Ch1cago:
Chicago Linguistic Socicty. Pp. 292-305. In tbe ordinary scnse in which we say that words like chair and tableare
ABOUT furniturc, hedges are words about language and speech. Therc
M~~~~·. ~~~~:~r~~~::C:r natural objects. Annual Rcview of Psychology 32:89- is nothing remarkable in this; language is part of our environment, and
115. we have words about most things in our environment. T he linguistically
Oc~~7~-~:~heE~niversallty of conversationnl implicatures. Langunge in Society interesting aspect of hedges is that, although they are about language, they
are not cxactly used to talk about languagc as we would say that chair
5(1):67-80.
and tableare used to talk about furniturc or, for example, gerund and
p¡~~~i. ~he Moral Judgement of the Child. M. Gabain, trans. New York: Har· entailment are uscd total k about language. Whcn we use a word like chair
court Brace. or table or gerund or entailment, chairs, tables, gerunds, and entailments
Rappapor.· R. _A. d lies In lntcrnationa1 Yenrbook for thc Sociology of do not beco me ipso jacto part of what is said. With hedges it is different;
19~~ow:~~r;:es~ Reli~ion, Vol. 10, G. Dux, cd. Frciburg, Germany: when wc use a hedge like loosely speaking, the notion of "loose speech"
Westdeutscher Verlag. Pp. 75-104. which this expression invokes becomes part of the combinatoria! seman-
tics of the sentence and uttcrance in which it occurs. A familiar (if prob-
Rols~~s. ~rinciples of categori:z.ation. In Cognition and Categori~tionpE. ~~c8h ably vacuous) combinatoria! semantic rule is
and B. Lloyd, eds. Hillsdale, N.J.: Lawrence Erlbaum Assocsates. p. .
k (SR) If adjectivc a denotes class A and noun n denotes class N,
S~~9.JSpeech Acts. Cambridge, England: Cambridge University Press. then the denotation of the expression an is the intersec-
Thomason, N. R. . d U blished Ph D disserta- tion of the classes A and N.
1983. Why we have theconcept of'he' that we o. npu . .
tion. University of Oregon. 1 wish to clairn that the notion of "loose speech" is part of the combinatoria!
semantics of sentences containing the expression loosely speaking in the
same way in which the notion of class intersection is claimed by proponents
of (SR) to be part of the combinatoria! semantics of an expression like
red chair.
A hedged sentence, whcn uttered, ofcen contains a comment on itself
or on its utterance or on sorne part thereof. For example, when someone
says, Loosely speaking France is hexagonal, part of what they have uttered
is a certain kind of commcnl on thc locution France is hexagonal. In this
son of metalinguistic commcnt, thc words that are the subject of the com-
ment occur both in their familiar role as part of the linguistic stream and
in a theoretically un familiar role as part of the world the utterance is abo ut.
Such metalinguistic reference seems unaccounted for (and perhaps unac-
countable for) in standard thcories of semantics that are based on a context-
free, recursive definition of truth for sentences, and in which linguistic